Zero-to-missing preprocessing pipeline and baseline comparison

Build a preprocessing pipeline that converts selected zeros to missing, imputes values, scales features, and trains Logistic Regression.
Compare performance against plain baseline.

Start in the repository's Jupyter Notebook by locating the current plain Logistic Regression baseline and the data-preprocessing flow. Compare the baseline with the zero-to-missing, imputation, and scaling pipeline; done means both performance results are available for comparison.

Written by the indexing model from the issue text.